Joe O’Donnell honed his stock-researching skills over a 13-year hedge-fund career, mining financial statements, securities filings and other data for hidden signals.
Now, he is trying to make much of what he did in his past job obsolete.
O’Donnell’s startup, Canary Data, makes artificial-intelligence software for professional analysts and investors that can perform tasks in hours that used to take him days or weeks. The three-year-old company is jumping into an AI arms race that has the potential to wipe out jobs across Wall Street while possibly supercharging others. The efforts are fueling anxieties across the banking and asset-management industries, where legions of professionals spend much of their days retrieving and synthesizing information.
Among Canary’s features: flagging companies with signs of fraudulent accounting or those with regulatory investigations that have yet to be disclosed. A new Canary AI agent, named Stanley after the investor Stan Druckenmiller, scans the market and its data sources for cyclical stocks that appear to be priced too high or too low, evaluating and writing up ideas in chart-filled memos.
Others are rushing to be the AI supplier for various corners of the industry, including Anthropic and OpenAI; established financial-software makers such as AlphaSense; and Canary’s fellow fintech startups.
Worries about AI-driven job losses in finance are on the rise, especially among younger workers. A recent survey of hedge-fund investment professionals found a “meaningful portion” expressed concern about whether their roles would survive in the long term, according to the recruiting firm Odyssey Search Partners.
Canary recently raised a new round of venture capital from Tiger Global Management, where O’Donnell previously worked as a partner, and Feroz Dewan’s Arena Holdings. Its customers include Tiger and the tech-focused Flight Deck Capital.
Canary charges annual prices ranging from the midfive to the midsix figures, depending on what services customers use, O’Donnell said.
There are plenty of things analysts do that Canary can’t. Humans are still better than AI at securing scarce access to company management teams and quizzing them in-person on granular aspects of the business.
But AI’s progress at handling Wall Street tasks is accelerating. Citadel Chief Executive Ken Griffin went from dismissing AI-generated work as superficial “garbage” in January at the World Economic Forum meeting in Davos, Switzerland, to praising the “step change” in AI during a recent talk at Stanford University’s business school.
O’Donnell tailored early versions of Canary to recreate parts of his past workflow, aiming to appeal to short sellers, his specialty at Tiger. It prepared risk summaries on stocks that factored in such things as related-party transactions and past misdeeds among executives or subtle changes to accounting policies.
As AI models got more powerful, he added more features and started targeting stock pickers, too. Canary now analyzes the predictive value of purchases or sales of shares from a given company insider.
It also has a new product that can generate fleshed-out debates featuring bull and bear cases for individual stocks, as well as custom research plans with suggestions on experts to interview or customer surveys to commission.
The name of the product: Superanalyst.
Write to Peter Rudegeair at peter.rudegeair@wsj.com